Fast Edge-Preserving Envelope for Dynamic Range Compression

High dynamic range scenes usually associate with abrupt illumination changes. In order to characterize such discontinuities in illumination, an algorithm for edge preserving image envelope estimation is proposed in this paper. The computational procedure follows the fast iterative approximate Gaussian envelope for Retinex theory, which is proposed by McCann and Sobel. Inspired by bilateral filtering, a contrast weight function is then introduced to preserve edges when computing the image envelope. The proposed envelope estimator is then adopted in dynamic range compression. Fine local contrast and low total dynamic range can be simultaneously obtained using the proposed algorithm. Meanwhile, the undesirable artifacts are avoided. Moreover, the algorithm inherits the high efficiency of McCann-Sobel Retinex algorithm.